Regression-Based Valuation
Regression-based valuation determines a fair multiple for a stock by running a statistical regression of price multiples (such as P/E) against fundamental drivers (such as earnings growth, ROCE, or dividend payout) across a peer set, generating a predicted multiple that reflects what the market pays for each unit of quality.
A simple comps analysis says company A's sector trades at a median of 25x P/E. Regression-based valuation goes further by asking: how much of the variation in P/E across the sector is explained by differences in growth rates, return on equity, dividend payout, and leverage? Once this relationship is quantified, a predicted P/E can be derived for a specific company based on its own fundamentals.
The most common form involves regressing P/E ratios (or EV/EBITDA) of 15 to 40 companies in a sector against their expected earnings growth rate (g), return on equity (ROE), and optionally dividend payout ratio (d). The result is an equation such as: Predicted P/E = 8.2 + 1.3(g) + 0.4(ROE). For a company with expected EPS growth of 18 percent and ROE of 22 percent, the model predicts a P/E of 8.2 + (1.3 × 18) + (0.4 × 22) = 40.0x. If it currently trades at 30x, the model implies it is undervalued relative to peers on these metrics.
Damodaran's PEG ratio approach is a simplified regression-based method: divide P/E by expected earnings growth to get the PEG ratio. A PEG of 1 is considered fair value in many markets, though this rule of thumb was developed in US market conditions and may not directly translate to Indian valuations.
The advantage of the regression approach is that it explicitly accounts for the growth-quality tradeoff rather than comparing raw multiples across companies with different fundamental profiles. The disadvantage is that it is only as good as the peer set and the time period chosen. Running the regression during a market peak will embed overvaluation into the 'fair' multiple.
This technique is most commonly used by institutional sell-side analysts in sector initiations, particularly for banking and financial services (regressing P/B against ROE), consumer staples (P/E against growth and dividend yield), and IT services (P/E or EV/EBITDA against revenue growth and margin). It requires multiple data points so it is not useful for unique businesses without listed peers.